I read a good piece last week by Hannah Ford over at Owner Toolkit, listing five businesses she thinks will do well over the next decade. Number four on her list was AI workflow and operations consulting. She's right, and I've been circling the same conclusion from the other direction for about a year.
One line in her piece stopped me though. In a parenthetical about how little hard data exists on this, she notes that AI is "predicted replace 1 trillion jobs."
A trillion is a thousand billion. Including you, , there are about 8.3 billion people alive. So that number (1,000,000,000,000) is roughly 120 times the entire human population.
That's not a knock on Hannah. She flagged the data gap herself, and that figure is floating around everywhere right now. It is the story. This category is so young that the numbers being passed around are off by two orders of magnitude and nobody blinks. When a market is that unformed, the person who shows up with a straight answer gets paid.
So here's a straight answer.
What AI workflow and operations consulting actually is
AI workflow and operations consulting is the work of examining how a specific company gets things done, finding the manual work that costs the most time or money, redesigning that process, and putting AI-assisted systems in place to run it, then training the team and measuring what changed.
The important word is operations. It's an ops job that happens to use AI, not an AI job that happens to touch ops. The consultant's product is a business that runs differently. The software is incidental.
Quick facts
What it is
Process discovery, redesign, automation build, team training, measurement
What it isn't
ChatGPT training seminars, tool resale, prompt libraries
Who buys it
Businesses with 5 to 50 people and real manual volume; solopreneurs buried in admin
Why now
The tools finally work, and almost nobody has mapped their own processes
Typical shapes
Paid audit, fixed-scope single workflow, or monthly retainer
Closest hard data
Management analysts, projected to grow 9% from 2024 to 2034 vs 3% for all occupations (BLS)
That BLS number is the best real figure we have, and it doesn't account for AI at all. It's the floor, not the ceiling.
The part everyone gets wrong
Most people hear "AI consultant" and picture someone running a lunch-and-learn on ChatGPT.
That's not the job. That's a webinar.
The job looks like this:
Watch how the work actually gets done, not how the org chart says it gets done
Find the expensive manual steps
Write the process down, often for the first time in the company's history
Decide what's worth automating and what isn't
Build it
Train the humans who now have to work differently
Measure the hours or dollars it gave back
Fix it when it breaks, because it will
Steps one through three are where the value is, and they involve no AI whatsoever. They involve a person with a notebook asking your operations manager to walk them through a Tuesday.
If a consultant leads with the tool -- "we'll get you on n8n," "we'll set up ChatGPT Teams," "we build custom GPTs" -- before they've watched you work, they're selling software. You can buy software yourself. It's cheaper.
Why it's a full-time job
Here's the thing that makes this a career and not a weekend project: every business is a snowflake.
There is no template. There is no playbook you buy once and apply to client after client. Every company has a process that exists only in one person's head, has three undocumented exceptions, and depends on a spreadsheet somebody's cousin built in 2019.
The consultant's real work is discovery: reconstructing a process nobody has ever written down, from the people who do it without thinking about it. That work does not compress. It does not scale into a product. It takes as long as it takes, at every single client, forever.
Which is exactly why it's a job.
Let me show you what I mean with my own business.
I publish a daily AI news briefing. Research, story selection, writing, illustration, voiceover, publishing, and social distribution, every weekday. If a person did all of that by hand it's four to five hours a day. Call it twenty to twenty-five hours a week, a half-time employee's entire workload, gone.
It runs on a schedule now. But building it took weeks, and almost none of that time was spent on the AI. The AI part was the easy part. The months went into figuring out my particular sequence: which sources are worth reading, what makes a story worth including, how the audio has to be structured, where the images go, what order things have to happen in so nothing publishes half-finished.
None of that was transferable from anyone else's setup. It couldn't be. It was mine.
The same was true on the OutSnapped side -- client communication, proposals, follow-up, scheduling -- and again on the software side with Pictor, where onboarding and support have their own logic. Three sets of workflows, three completely different maps, one company.
Now multiply that by every small business in your city. That's the job. That's why it's full time.
Most of your processes shouldn't be automated
This is the part that separates people who've done this from people selling it.
A good audit tells you what to leave alone.
Some processes are low volume and high judgment, and automating them costs more than doing them. Some are the actual relationship with your customer and should stay human on purpose. Some are so broken that automating them just produces bad output faster. And some are about to change anyway, so building around them now is a waste.
If someone maps twelve of your workflows and comes back recommending you automate all twelve, they didn't do an audit. They did a quote.
The right answer at most small businesses is two or three processes, done properly, that give somebody their afternoon back. Then you live with it for a quarter and decide what's next.
How to hire one
Five questions that sort real operators from prompt tourists.
"What would you look at first?" If the answer names a tool, that’s your answer. It should name a process or a person.
"What have you automated in your own business?" Anyone doing this well has done it to themselves first. Ask what broke.
"What would you tell me not to automate?" They should have an opinion, and it should be specific to what you’ve already told them.
"How will we know it worked?" Hours per week, error rate, response time, something countable. If the answer is "efficiency," keep looking.
"What happens when it breaks?" Because it breaks. Models change, a vendor moves an API, a login expires. Somebody has to own that.
What it costs
There's no single number, and anyone who quotes you one before understanding your business is guessing.
The shape of the engagement depends on the size of the company, how many people the change touches, and how many separate workflows are in scope. Broadly it lands in one of three places.
A paid audit. A week or two of mapping, ending in a prioritized list of what to fix and what to leave alone. Build work quoted separately. Good when you don’t know where the time is going.
Fixed-scope build. One workflow, one price, one deadline. "Automate our intake." Good when you already know exactly what’s bleeding.
Monthly retainer. Someone owns AI operations for you on an ongoing basis: builds, maintains, trains, keeps it running. Good once you have more than one system live and something has already broken once.
A solopreneur usually needs the second one. A twenty-person company usually starts with the first and ends up at the third.
Where this goes
The reason I'm writing this is that I've spent the last year doing this job for my own companies, and I'm starting to do it for other people's.
If you run a business between five and fifty people and there's a process eating hours every week that you know shouldn't, hit reply on this email and tell me two things: what your business does, and which task you'd hand off first if you could. I read every one of these myself.
And if you're a solopreneur looking at the same problem with a smaller budget, reply anyway. The answer might just be a thirty-minute conversation and a suggestion, which is a fine outcome.
The trillion-jobs number is nonsense. The work underneath it is not.
Further reading: Stop Running AI Agents Like Freelancers. Give Them a Company. -- what actually goes wrong once you have systems running unattended.
This started with Hannah Ford's 5 Smart Businesses That Will Thrive in the Next 10 Years at Owner Toolkit.